Conventionally, a learning control of fuel injection amount is performed after a vehicle was assembled or after the vehicle was shipped, in order to correct a variation (dispersion) in the fuel injection amount of an engine.
Such learning control method is disclosed, for example, in Japanese Patent Application Laid-open No. 2004-11511. Such learning control method involves finding an idle fuel injection amount (for example, 5 mm3) by taking into account the life cycle of a vehicle or a variation of amount in each vehicles, and reading this idle fuel injection amount in an engine control unit in advance, before learning control of the fuel injection amount is started. The engine is then operated in an idle mode and learning of the fuel injection amount is started. Where the engine speed is higher than the idle speed (for example, 800 rpm) during this learning, a learning value (for example, a fuel injection period correction value) is subtracted and a learning value at a time the idle speed is assumed is saved in the engine control unit as “a learning completion value at which the idle injection amount is correctly injected”.